mushin
mushin copied to clipboard
c0dearm
ML examples
Hi there!
This is a really cool library and I'm excited to try it out (especially after spending several hours fighting my environment to get tch-rs working >_> - still haven't, suspect it's easier to rewrite in mushin instead!)
That being said, I'm still a bit of a novice at ML, and would appreciate some applied examples to see how you might port e.g. one of the PyTorch examples over. (I'm particularly interested in making a GAN or a VAE, but don't let that influence your choice!)
No rush on this, but it'd be nice to have, especially for other people who'd like to better understand how to use the library 🙂
(Unrelated: is there any kind of roadmap for what you're planning on implementing next / would like to see implemented?)
Hi! :smile:
It is good you ask about this because actually I am currently working on writing a MNIST digit recognition based on this.
So it should be ready any time soon. I have one stopper which I am trying to solve, which is that when writing the model and chaining multiple layers, I get into a trait bound hell, something similar to what is discussed in this Reddit thread. Of course I want to make it easy for final library users to write their own neural network models so at the same time I am writing the MNIST example I am also doing some refactors of the library to provide the best experience possible.
But yeah, answer is, I am working on it. For now I have only planned to write the MNIST example but I don't discard implementing a GAN afterwards too.
Regarding the Roadmap, for now I don't have very long term planning. But I do have a short time one that also helps me to organize my work, I use a GitHub project for this. It is public, so you should be able to see the status of different tasks. For now the goal is to have this MNIST example fully working and release version 0.6.0 with it.
